What kind of illumination and how much do you need to make a hyperspectral imaging application? We know you need much more than when using an RGB or black-and-white camera.

In the visible wavelength range, LEDs are already an option for halogen light, but in the near-infrared region, you need a halogen lamp, although LEDs are soon a viable option.

But when you are using a halogen lamp, how much do you need to illuminate your sample and get reasonable results?

I have a simple rule of thumb, so let’s look at that.

So, what is the illumination power needed for FX17 in the near-infrared region? In practice, this is a very difficult question, and in most cases, I make tests for this.

But I made a very simple rule of thumb that gives the approximate power needed, and the equation is actually simple.

We have a power as a constant. Times r squared, divided by integration time.

So, in this, the constant is ten watts, second divided by square meter, the t is the integration time used, and R is the distance from the illumination line to the target.

To explain this, I have two examples.

First, we assume that the distance from the illumination to the surface is something like 25 centimeters, and the integration time is 0.01 seconds.

Simply calculating, based on this, we get that the power needed is 60 watts.

And, of course, when we are illuminating a line, we should divide this power so that the whole line of 20 to 30 centimeters is covered evenly.

And we can get that one can achieve this illumination by using six halogen lamps with a power of ten watts each, maybe 15 or 20 watts each, and two times a 45-degree configuration.

In practice, this is what we have in the 40×20 scanners, where the illumination is about 30 cm in width and made of three halogen lamps
and actually, three halogen lamps in two times at a 45-degree assembly.

My next example is a bigger system with a distance from the illumination to the surface of about 50 centimeters, and our integration time is 5 milliseconds. Again, putting this to the equation, we get that the power needed is 500 watts, which is actually close to ten times more,
and doubling the distance from the light source to the surface and only reducing the integration to half increases the need for illumination really, really quickly.

And this actually is the case when we are looking at the bigger systems
where that line width is something like 70 centimeters up to one meter.

The lessons we learn from this is that the illumination should actually be as close as possible to the sample surface, and one should use as long integration time as possible. Of course, this is a compromise between sample resolution, speed of the conveyor belt, and so on.

But to reduce the need for illumination, we should think about the distance of our light source or to the surface that we need to measure.